Comparison of Performance Records and National Breeding Values

J. Dairy Sci. 87:2709–2719  American Dairy Science Association, 2004.

Comparison of Performance Records and National Breeding Values as Input into International Genetic Evaluation W. F. Fikse Interbull Centre, Department of Animal Breeding and Genetics, Swedish University of Agricultural Sciences, S-750 07, Uppsala, Sweden

ABSTRACT The purpose of this investigation was to compare accuracy and precision of variance components and breeding values for international genetic evaluations based on national breeding values or animal performance records. A conventional progeny test scheme was simulated for 3 countries. True breeding values and observations were generated specific to production environments. Two production environments were considered, and both balanced and unbalanced distribution of production environments over countries were considered. True breeding values for both production environments were generated as bivariate normal deviates, and low (0.70) and high (0.90) genetic correlations between performance in production environments were considered. Each cow had an observation in one country only. Performance records were generated as the sum of the true breeding value, a contemporary group effect, and a random residual. Eight generations of data were simulated, and the entire simulated data set was used to compare 3 methods for international genetic evaluation: 1) multiple-trait across-country evaluation based on national predicted breeding values of bulls (MACE), 2) international genetic evaluation across country using performance records, and 3) international genetic evaluation across production environment using performance records. Estimated genetic parameters were biased for all models in this study. Genetic correlations between countries were generally more biased for MACE than for the across-country analyses using performance records. Bias in within-country genetic variances was smaller for MACE. Even genetic parameters obtained with the international evaluation across production environment using performance records were biased, despite the fact that this model was closest to the true, simulated model. The root mean square error of predicted breeding values was similar between models for most of the situations considered. The difference be-

Received August 6, 2002. Accepted February 3, 2004. Corresponding author: W. F. Fikse; e-mail: Freddy.Fikse@hgen. slu.se.

tween models was largest when the distribution of production environments over countries was unbalanced and the genetic correlation between performance in production environments was low (0.70). Using breeding values obtained with the across-production environment international genetic evaluation based on performance records will increase the response to selection. (Key words: simulation, borderless evaluation, multiple-trait across-country evaluation) Abbreviation key: GAMcou = international evaluation on country basis using performance records, GAMenv = international evaluation on production environment basis using performance records, MACE = multiple-trait across-country evaluation, NAMcou = national evaluation using performance records, PBV = predicted breeding value, RMSE = root mean square error, TBV = true breeding value. INTRODUCTION Genetic improvement and genetic evaluation programs for dairy cattle have been primarily organized within country. However, a more recent development is the growing importance of international links between breeding programs (McGuirk, 1998). Selecting bulls across countries can enhance genetic progress considerably when countries have similar or identical breeding goals (Banos and Smith, 1991). To realize this progress, breeding value predictions should be similar in all countries. Differences in genetic evaluation systems influence genetic correlations between countries (e.g., Banos et al., 1992; Emanuelson et al., 1999) and may hamper genetic progress (Lohuis and Dekkers, 1998). Therefore, Lohuis and Dekkers (1998) suggested consideration of borderless evaluations that standardize data collection and evaluation across borders and in which traits are based on production environments instead of country boundaries. Weigel and Rekaya (2000) and Zwald et al. (2001), both working with field data, clustered herds across country borders according to information on production systems. The lowest genetic correlation between performance in production environments was 0.81 (Weigel

2709

2710

FIKSE

and Rekaya, 2000) and 0.59 (Zwald et al., 2001), indicating the presence of genotype × environment interaction. Furthermore, the number of production environments was smaller than the number of countries, and boundaries of production environments and countries did not overlap perfectly. Consequences of similar situations as described before for the optimum design of breeding programs and genetic improvement were investigated by Banos and Smith (1991) and Lohuis and Dekkers (1998), using deterministic simulations. However, little is known about the gain in precision of predicted breeding values (PBV) when international genetic evaluations are borderless. The aim of this study was to investigate gain in precision and accuracy of estimated variance components and predicted genetic merit when individual performance records can be used instead of national breeding values and when international evaluation is across production environment rather than country. MATERIALS AND METHODS Three countries and 2 production environments were simulated. Animals were generated within country, and observations were within production environment. That is, the breeding programs were run within countries, but true breeding values (TBV) and observations were generated specific to production environments. Performance in each production environment was treated as a separate biological trait. Unit of time was generation, and 8 discrete generations were simulated in addition to an unselected base generation. The simulation was repeated 15 times. Generated data were subsequently used as input to 3 different methods of international genetic evaluation. Simulation of Data The breeding program was a conventional progeny test scheme. Countries A and C were of equal size and comprised 20,000 cows per generation. Country B had only one-half as many cows per generation. In each generation, 200 young bulls were progeny-tested in countries A and C and 100 bulls were progeny-tested in country B, of which the best 10 and 5%, respectively, were used as proven bulls and sire of sons in the next generation. Progeny group size for young bulls was normally distributed around a mean of 80 (SD = 5). Proven bulls that had passed the progeny test received an additional 200 progeny in the next generation. One hundred elite dams in countries A and C and 50 elite dams in country B were selected as dams of sons, and each elite dam was mated to one sire of sons to produce 2 full brothers. Each cow was mated to either a proven or a Journal of Dairy Science Vol. 87, No. 8, 2004

Figure 1. Distribution of production environments over countries.

young bull to produce one female offspring for the next generation. Mating of selected animals was random. Selection of animals to produce the next generation was done within country, regardless of production environment. The criterion for selection was national breeding values predicted within country. Sires of sons and proven bulls that had been selected within country were exchanged from generation 3 forward. Country B imported 30 and 20% male genetic material from countries A and C, respectively. Each of countries A and C imported 25% male genetic material from the other. Two different distributions of environments over countries were considered (Figure 1). In the first situation (balanced), 50% of the observations in each country were made in each production environment. In the second situation (unbalanced), 90% of the observations in countries A and B and 10% of the observations in country C were made in the first production environment. Performance in production environments was treated as genetically distinct, although correlated, traits. True breeding values for first lactation production in both production environments were generated as the sum of the parent average and a Mendelian sampling deviation. This Mendelian sampling deviation was drawn from a multivariate normal distribution and reflected the genetic covariance structure among both traits as well as the inbreeding coefficients of the parents (Fikse and Banos, 2001). Genetic variance and heritability were 210 and 0.30, respectively, for both traits. Genetic correlation between performance in both production environments was 0.70 or 0.90. Observations were generated as the sum of a contemporary group effect, the TBV and a residual. Contempo-

2711

INPUT VARIABLE IN INTERNATIONAL GENETIC EVALUATION

rary group size was, on average, 20 (SD = 5) with a minimum size of 5. Contemporary group effects were drawn from a normal distribution with mean zero and variance equal to 10% of the total phenotypic variance. Contemporary groups were nested within country and production environment. Cows were assigned to contemporary groups such that daughter-dam pairs were likely to produce in the same production environment. Breeding values were predicted for each country separately in each generation after observations had been generated. A model was used that included fixed contemporary and genetic group effects as well as a random animal effect. Genetic groups for phantom parents (of base animals and imported animals) were formed on the bases of country of origin and generation of birth. Simulated variance components were used in the national genetic evaluation. The mixed model equations for breeding value prediction were set up according to the implicit representation method, which reduces memory requirements by taking advantage of the occurrence of repeated blocks in the equation system (Tier and Graser, 1991). The mixed model equations were solved with a pre-condition conjugate gradient algorithm (Strande´n and Lidauer, 1999). The pre-conditioner matrix was block diagonal, and, for fixed effects, the complete diagonal block of the left hand side of the mixed models equations was used. For animals, the diagonal blocks pertaining to all equations of an animal were used. Solutions of the mixed model equations were assumed converged when the average relative difference between consecutive rounds was